New ways to support struggling readers: Evidence from some

recent interventionsDyslexia Guild Annual Conference

June 20th 2019Robert Savage

University College London

Institute of Education

Kristina Maiorino

Faculty of Education

McGill University, Montreal, Quebec

Canada

1

Rauno Parrila

George Georgiou and Kristy Dunn

University of Alberta

Edmonton

Alberta

Canada

MacQuarie University Australia

The Importance of Early

Intervention

• Early intervention is crucial to the prevention of reading difficulties and raising literacy standards.

• 1 in 5 children struggle with literacy

• Up to two thirds of poor readers can learn to read at age-expected levels with appropriate early intervention (e.g. Savage, Carless, & Erten, 2009)

• The principled assessment of evidence-based quality teaching allows the exclusion of ‘environmental deficit’ in candidate cases of developmental dyslexia

2

Current Best Practices • Focus on systematic teaching of phonics – the

assembly of pronunciations of words from component

speech sounds (e.g. McArthur et al., 2012)

• Teach letter-sound knowledge (grapheme to—

phoneme correspondences) and phonological

awareness

• Expose children to books e.g. shared book reading

• Teach vocabulary

• Teach for and with practices to assist comprehension

• Quality teaching differentiated to needs

3

Design: 4 sub-studies with teacher

training and direct student support

• Intervention 1: Kindergarten o Parental early involvement

interventions with technology.

• Intervention 2: Grade 1 o A) Direct Mapping and Set for

Variability(DMSfV).

o B) Current and Best Practices (CBP)

• Intervention 3: Grade 2 o Spelling Intervention using the

Simplicity Principle.

o Vocabulary-taught control

• Intervention 4: Grade 3 o Morphological Awareness and

Vocabulary Intervention.

4

Decoding

Spelling

Vocabulary

New theory-driven interventions

• One view is that reading fluency is the use of

accurate automated Grapheme Phoneme

(GPC) and other letter-string data to access

word knowledge within a connectionist

network (Ehri, 2015)

• So: opportunities to automate GPC links in

words richly represented in real texts likely

to be important

• Consider the words ‘w’ ‘a’’s’’p’ and

• ‘b’ ‘e’’l’’t’

5

Set-for-Variability• English is opaque or ‘deep’ (Seymour et al., 2003)

• So: strategies for dealing with depth (especially

variable vowels) are important

• A ‘set for variability’ Elbro et al., 2012; Tunmer &

Chapman, 2012)

• ESSENTIALLY – the idea that we need to do more

than apply phonic rules to access stored

pronunciations (and this MAY be particularly true in

a deep orthography)

6

Set-for-VariabilityElbro et al. (2012, Elbro & de Jong, 2017)

• Decoding is:

• A) phonemic assembly e.g. ‘d’-’o’ –’g’.

• B) matching a ‘spelling pronunciation’ to a

known word in their lexicon … e.g. ‘dog’

• A ‘2 process model’ of assembly

• The first part has historically been emphasized in

education

Set-for-VariabilitySfV predicts growth in regular and exception word

reading (Elbro et al., 2012).

SfV is associated with vocabulary knowledge and

reading (Tunmer & Chapman, 2012)

What happens if we teach decoding as:

A ‘2 process model’ of assembly

• Does it measurably improve decoding, word

and sentence reading and spelling over

common /best practices?

MethodologyDesign

• A CONSORT quality dual site 2-arm cluster RCT

• Tier 1 and Tier 2 Pre-post-test design and grade 2

delayed post-test

Participants

• Tier 1

• Quebec, Alberta n = 429 grade 1 students

• Nested in 50+ classrooms in 26 schools teachers

• Tier 2

• All at-risk participants after fall semester (n= 201)

9

Winter tier 2 Intervention

• Mid-test: Re-screened and identified ‘at-risk’ = all children below the 30th % -ile on WRAT word reading test (n = 201)

• BY school groups of 3-4 children in one of 2 interventions:

• 1) A novel RtI phonics / word reading intervention called DMSFV: Direct Mapping and Set for Variability

• 2) A standardized ‘Common or Best Practices’ (CBP) phonics / word reading intervention

• For winter semester (average time per child = 10-11 hours) 30 minutes 2-3 times per week outside

10

Intervention 1: DMSfVThe DMSfV (Direct Mapping and Set for Variability) intervention:

• Systematic synthetic phonics: Taught phonic rules (espvariable multiple vowels e.g. ‘ou’ and ‘magic e’ rule

• Taught SfV principles and to substitute (‘flip’) GPCs from given rules to find a word in lexicon that makes sense

• Trained children to use either alternative phonics rules when decoding did not work (e.g. ‘wasp’, ‘shoulder’ regularized)

• Shared book reading with text for each session embodying the specific taught graphemes or exception words of that day

• ALL differentiated for the reading levels of children and sense of ‘playfulness / games / active learning

11

Intervention 1: DMSfV• We first taught the concept that phonic rules do not

always yield a clear pronunciation of words and that a second process (a ‘strategy’) is needed.

• A staged 5-step plan for variable vowels:

1. Children blend phonemes of a letter string, looking for and applying well-taught phonic rules.

2. Children evaluate their first attempt to synthesize a pronunciation: ‘Is this a word I know?’)

3. If no, children then replace the vowel with an alternative vowel pronunciation they know

4. Children synthesise this revised phoneme string

5. Children re-evaluate this blended string using the same reflective lexical decision process as above.

12

High densities of taught units in shared book reading

13

Intervention 2: CBP The Common and Best Practices phonics / reading intervention had the following characteristics:

• Equal time spent on ‘dual foundations’ of regular phonic rules (synthetic phonics) and high frequency exception ‘sight words’

• Taught common phonic rules for common words (e.g. some vowels and ‘magic e’) BUT DID NOT teach production of words from variable rules

• DID NOT train children to use either alternative phonics rules or wider context when decoding did not work (e.g. ‘wasp’, ‘shoulder’ regularized)

• Shared book reading DID NOT embody the specific taught graphemes or exception words of the day (but was differentiated for reading levels of children)

• Did have same differentiation and ‘playfulness’ 14

Results: WRAT Word Reading

15

55

60

65

70

75

80

85

90

95

100

DMSfV CBP

Pre grade 1

Mid grade 1

End grade 1

Start grade 2

Results: WJ Pseudoword decoding

16

55

65

75

85

95

105

115

DMSfV CBP

Pre grade 1

Mid grade 1

End grade 1

Start grade 2

Results: GRADE Sentence Comprehension (stanine)

0

1

2

3

4

5

6

7

8

End grade 1 Start grade 2

DMSfV

CBP

17

Conclusions: Grade 1 study • The ‘tier 2’ reading interventions did show

effects even after documented good regular tier 1 teaching

• BUT The DMSfV approach was far the more successful of the two for word reading and phonological awareness

• CBP is no guarantee of success

• Medium effect sizes for interventions evident at delayed post-test for DMSfV intervention

• Other data suggest it transfers into French

18

Intervening in Grade 2 why phonic skills? (Again)

Phonics is based on 2 skills:

• Letter-sound knowledge and phonemic awareness

• Both need to be supported and

in-place for children to be able to

‘sound out’ or ‘decode’ or use

‘phonics’

Robert Savage 19

The effects of the spelling system: Seymour et al.

• Scottish and other European children do not differ

on basic letter sounds > 90% accuracy

• Scottish children know 1/3 of the matched content

and function words of most European counterparts

(except Denmark)

• Europe and Scandinavia: 90% accuracy (by P1)

• Denmark: P1 72 % P2 : 92%

• Scotland: P1 34 % P2 : 76%

20

Phonics programs for Grade 1

Jolly phonics:

Robert Munsch

22

Example (The paperbag princess)

Elizabeth gra**** the **ocker and b***** on *** d***

ag**n.

*** dragon stuck hi* n*** out o* *** d*** and s**d, “Go

aw**, I l*v* to **t prin*ess*s, but I h*** alr**d* **t** a

**ole cas**e tod**. I am a ver* b*s* dragon. C*m*

back t*mor***.”

“Wait,” shout** Elizabeth. “Is it tr** that *** *r* **

smartest and f**r*est dragon in *** *h*l w**ld?”

“Yes,” s**d *** dragon.

Example (The paperbag princess)

Elizabeth gra**** the **ocker and b***** on the

d*** ag**n.

The dragon stuck hi* n*** out o* the d*** and s**d,

“Go aw**, I l*v* to **t prin*ess*s, but I h*** alr**d*

**t** a **ole cas**e tod**. I am a ver* b*s*

dragon. C*m* back t*mor***.”

“Wait,” shout** Elizabeth. “Is it tr** that *** *r* the

smartest and f**rcest dragon in the *h*l w**ld?”

“Yes,” s**d the dragon.

Example

Elizabeth grabbed the knocker and banged on the

door again.

The dragon stuck his nose out of the door and said,

“Go away, I love to eat princesses, but I have

already eaten a whole castle today. I am a very

busy dragon. Come back tomorrow.”

“Wait,” shouted Elizabeth. “Is it true that you are the

smartest and fiercest dragon in the whole world?”

“Yes,” said the dragon.

The Simplicity Principle

Background• “The theory of optimal instruction states that there is

an optimal amount of information to teach that will

lead to maximum generalisation” (Solity & Vousden, 2009,

p.9)

• What is the optimal amount of information to teach

in reading?

Background• Analysis of 685 contemporary children’s books:

o 100 most frequent English words account for approximately 50% of all

word

• the next 50 most frequent words account for a significantly lower

proportion of word tokens in children’s books

o 64 most frequent GPCs (out of 461 in English) would enable children to

read more than 60% of all word types in children’s books

Proportion of monosyllabic word tokens that can be read as a functionof the most frequent grapheme-phoneme mappings (Vousden, 2008, p.262)

Most frequently occuringGPCs

Vousden and Solity (2011)Ranked List of Grapheme-Phoneme Mappings

“s” = /zz/ ed oa ow

sh ss dg ur

ee = /eeee/ th ou kn

a_e o_e wh gg

ch ai ed oi

pp aw ay i

ng ir or air

ck tch oo eer

ll ff th ore

i_e ar ow ear

ea = /ee/ igh qu etc…

A new Canadian analysis

• We obtained the 500 most frequently borrowed

young children’s books from all of Toronto district

public libraries – summer 2014!

• My team typed them all into a database and Dr

Solity’s team in the UK analyzed this list to provide a

specifically Canadian list of most frequent GPCs,

most common words, and their links to the most

popular Canadian real books

• We shared these maximally useful units with

teachers to aid their teaching and we used them in

the small group winter interventions too!

A pilot study (Chen & Savage, 2014)

• We randomly allocated 38 grade 2 students to a 9-

week 30 supplemental small group session

programme.

• We taught intervention and taught control

conditions.

• Intervention reading programme taught children

complex GPCs ordered by their frequency of

occurrence in children’s texts (a ‘simplicity

principle’).

• The other reading programme taught children

word usage.

tim

e

33

Example: Shared

reading and identifying GPCs

Elizabeth grabbed the knocker and banged on the

door again.

The dragon stuck his nose out of the door and said,

“Go away, I love to eat princesses, but I have

already eaten a whole castle today. I am a very

busy dragon. Come back tomorrow.”

“Wait,” shouted Elizabeth. “Is it true that you are the

smartest and fiercest dragon in the whole world?”

“Yes,” said the dragon.

Example

Elizabeth grabbed the knocker and banged on the

door again.

The dragon stuck his nose out of the door and said,

“Go away, I love to eat princesses, but I have

already eaten a whole castle today. I am a very

busy dragon. Come back tomorrow.”

“Wait,” shouted Elizabeth. “Is it true that you are the

smartest and fiercest dragon in the whole world?”

“Yes,” said the dragon.

A pilot study (Chen & Savage, 2014)

• Participants in the complex GPC group performed

significantly better at post tests

• Generally large value-added effect sizes (Cohen’s

d) at both by-participant and by-item for :

o spelling, d = 1.85, d = 1.16

o word recognition with words containing taught

GPCs, d = 0.96, d = 0.95

o word recognition, d = 0.79, d = 0.61

o reading motivation, d = 0.34, d = 0.56

Pan–Canadian replication study

• 2 Provinces in East and West of Canada with

around 150 at-risk students screened from 500.

• Well-matched quasi-experimental design with

• School level randomization

• Controls for demographics, regular teaching quality

• Language background, parent reports of difficulty

• Quality of training and delivery of interventions

• Early literacy and language abilities

• School literacy experiences

• Nested analysis of 50 classrooms

37

Replication study

• We find a significant interaction:

• Word, pseudoword, sentence reading,

spelling

• Simplicity was effective for those with higher

phoneme blending skills

• 30+ years of research shows that both GPC

and phoneme awareness training is

essential for improvements.

38

Some important caveats • 2 of 3 in this sample remained below

average on phonological awareness at

post-test (mean more than 2 SDs below

average)

• Average comprehension composite was SS

= 85 at post-test

• Clear impact on word-level skills but not a

‘magic bullet’ for comprehension or fluency

39

Grade 3 struggling readers

• Here struggling readers have unique needs

• They may have had several years of ongoing

support for phonics

• Do we give more of the same? (as in RtI?)

• They often have greater challenges in facing

reading and understanding meaningful text

• Should we thus support wider language needs?

• We thus contrast Simplicity (with phonic blending)

and:

• Structured Word Inquiry

40

Structured Word Inquiry • Assumes that English is morpho-phonemic:

• Venezky (1967) “the simple fact is that the present orthography is not merely a letter-to-sound system riddled with imperfections, but, instead, a more complex and more regular relationship wherein phoneme and morpheme share leading roles” (p. 77).

• The key phrase here is that “phoneme and morpheme share leading roles.”(Bowers & Bowers, 2018). Morpheme = smallest unit of meaning

• Consider e.g. ‘bomb’-’bombardment’, ‘nation’-’national’, ‘sign’- ‘signal’

41

The SWI approach • Focus on GPCs within a frame of morphology and etymology

• A morpheme is the smallest unit of meaning

• Roots (e.g. ‘happy’ and bound morphemes (essentially prefix and suffix units ‘-un’, ‘-ness’ etc

• Etymology is the study of the historical origins (and thus of connectedness) of words

• e.g. consider ‘ react’: Morphology clarifies the phonology.

• The morphological structure = < re + act>,

• This rules out <ea> as a digraph because graphemes never cross morphemic boundaries.

• More generally it explains why the word ‘action’ includes the <t> rather than the <sh> grapheme to represent the /∫/ of action).

42

The SWI approach

43

SWI and Simplicity contrasts

44

Georgiou, Savage, Dunn, Bowers, & Parrila (2019)

• Contrasted SWI and Simplicity

intervention models

• 48 English-speaking children with

reading difficulties randomly

assigned to one of:

• Simplicity Principle (SP),

• Structure Word Inquiry (SWI),

• No Intervention (Control). 45

Georgiou, Savage, Dunn, Bowers, & Parrila (2019)

• Interventions: 10 weeks, 3 times a week for 30 minutes by trained psychology graduate students.

• Pre-test, post-test, and delayed post-test on: phonological awareness, morphological awareness, reading (word reading, word attack, morphological relatedness), and spelling.

• Groups well matched at pre-test: attainment, home language, parent–reported developmental history, age and gender

46

Results: WRAT word reading

•

47

74

76

78

80

82

84

86

88

90

SWI Simplicity Control

pre-test

post-test

delayed post-test

Other results • Similar significant effects on a morphological

reading task

• BUT no main effects on the other ‘secondary’

outcome measures

• Across these other measures there was an

interaction effect

• i.e. the interventions all worked best for children

who started with high pre-test morphological ability

• Small sample suggests need for caution here (i.e.

more work is needed)

• Can trained teachers produce the same effects?

48

Summary • We have shown using sustained cross year interventions

that a range of theory- and data –driven practices can improve reading in at-risk samples at scale across regions in grades 1, 2, & 3.

• Effects are evident in overlapping samples across years and in grade 2 despite Grade 1 effects

• Effects suggest we learn to map complex GPCs best where we map them to texts and with conceptually-driven SfV strategies for making sense of such heuristics across all words.

• In English the complexity of the system in English necessitates extra ‘word level’ work with optimal GPCs in grade 2

• In Grade 3: preliminary data suggest that SWI helps best.

49